Lightweight and smart data fusion approaches for wearable devices of the Internet of Medical Things

Internet of Medical Things (IoMT) refers to a group of dedicated devices, i.e., wearable sensors attached to the human body for capturing the real-time data, specifically physiological signals, to develop an automated and technology-assisted smart healthcare framework. However, due to the miniaturized nature and their limited capabilities, data captured by these wearable devices need to be consistent, especially in terms of accuracy and precision. Therefore, data must be refined prior to the actual processing, which is a challenging task in a smart healthcare infrastructure. In the literature, various approaches have been reported to address this issue. However, most of these approaches are either domain-specific or extremely difficult to implement, especially in the resource-constrained environments of IoMT. In this paper, we present two-tier lightweight device and server-enabled data fusion approaches for wearable devices and their server modules, respectively. In these approaches, each device is bound to perform an in-node processing, i.e., local data fusion, on the sensed data values prior to initialization of the transmission process. For this purpose, each wearable device is required to use knowledge of the previously transmitted refined data values. Additionally, the proposed approach is pruned against abrupt changing scenarios, e.g. heart attack, where the frequency of sensed data values drastically fluctuates. Moreover, a non-metric-based data fusion, i.e., global data fusion, is presented to enhance the accuracy of sensed data by removing redundant data values, especially those collected by neighboring devices. Simulation results reflect the exceptional performance of these smart data fusion approaches, i.e., both local and global, in the context of numerous performance evaluation metrics, such as accuracy and precision ratio.